A language model can hand you an object that is perfectly shaped and entirely false. Every field is present, every type is right, the whole thing parses without complaint, and the value in the field your downstream logic depends on was invented. This is the failure mode that survives structured output. Enforcing a schema settles what shape the answer takes. It settles…
Two of the cheapest and most reliable ways to steer a model are to tell it who it is and to tell it what it knows. The first is identity: a role assignment, optionally extended into a full persona. The second is grounding: placing the specific facts a response should reason over directly into the prompt. Both are levers you pull…
Once you accept that a summarizer will drop the exact values a long session depends on, the fix is easy to describe and easy to half-build. Pull the facts out of each turn before compression runs, write them into a durable region the summarizer never touches, and the values survive while the narrative is free to shrink. Stated that way, a…
Inside a single agent, coordination is almost free. Everything the model has read, decided, or produced sits in one context, and any later step can reach back and use it without anyone arranging for that to happen. The moment work is split across separate agents, that convenience is gone and most people do not notice it has left. Each agent runs in its…
Dispatching work to run concurrently is the easy half of parallel execution, and it is the half everyone thinks about. A single control point splits into several branches, each branch runs at the same time as the others, and the total time drops from the sum of the branches to the length of the longest one. That part is close…
A retry is the most ordinary thing an autonomous system does with a failure, and the one a tool is least often built to survive. When a call comes back failed, the model may simply issue it again, on its own judgment, with less context than the engineer who wrote the tool and no way to know whether a second…
Retrying is the most reflexive move in error handling and the least examined. Something failed, so try it again. The reflex is defensible often enough that it survives without scrutiny, and that is exactly the problem. A retry is not a neutral safety net. It is a wager that the thing which just failed will succeed if attempted again, and that wager…
A tool definition is normally understood as a way to make something happen. You declare a function the model can call, it emits the arguments, and something on the other side runs. For pulling structured data out of text, invert that picture entirely. Define the tool, let the model call it, and let nothing on the far side ever run. The call itself is…
Every prompt you send a model is really two documents fused into one. There is the part that is identical on every request in a session, and the part that is different every time. The single most consequential decision in prompt architecture is where you draw the line between them, because that line determines what the model treats as…
Long conversations outgrow the window that holds them. A session that runs for hours, an agent that works a task across dozens of turns, a support thread that keeps accumulating, all of them eventually carry more history than the model can take in at once. The standard answer is to compress the older part of that history into a shorter…